Accountability chains in agentic systems: who is responsible when models act?
Legal, ethical, and operational frameworks for establishing clear accountability when autonomous AI agents make decisions with significant consequences.
Authors
K. Adeyemi, SOLON, 33 contributors
Published
2030
Citations
427
Overview
As AI systems become more autonomous, responsibility for their actions becomes murky: is it the model builder, the deploying organization, the operator, or the AI system itself? This research develops accountability frameworks tested across legal, financial, and regulatory domains.
Methodology
Legal analysis across 12 jurisdictions examining how liability is assigned in high-stakes AI deployments. Case studies of 15 significant AI failures and liability resolution. Interviews with 40+ legal experts, regulators, and affected parties. Framework testing with regulatory bodies.
Key Findings
Current legal frameworks are inadequate for agentic systems: case law assumes human decision-makers and struggle with autonomous systems. Most jurisdictions default to assigning accountability to either the organization operating the system or the model developer, but clear liability standards are absent in 11 of 12 examined jurisdictions.
Accountability frameworks require three layers: (1) technical accountability (ability to trace decisions and understand reasoning), (2) organizational accountability (governance structures and approval processes), (3) legal accountability (liability assignment and remediation). All three layers must be in place for systems to operate responsibly in high-stakes contexts.
Explicit accountability assignment (clearly defining who is responsible for what decision types, thresholds, and outcomes) reduces organizational AI risk by 71% and improves decision quality by 28% compared to ambiguous accountability. This suggests clarity about responsibility actually improves autonomous system performance.
International accountability frameworks are beginning to emerge through EU AI Act, proposed US rules, and industry best practices. Organizations adopting accountability frameworks proactively are better positioned than those waiting for regulatory mandates, experiencing 40% lower compliance costs when frameworks are eventually regulated.
Impact & Application
Informs policy development across multiple jurisdictions. Accountability frameworks adopted by 12+ major enterprises. Shapes emerging AI regulation and liability standards.
Contributors
Lead: Dr. Kofi Adeyemi (AI Governance school). Legal collaborators from leading firms (Allen & Company, Cooley, Latham & Watkins). Regulatory advisors from EU, US, and UK authorities.